|
Faculty of Engineering Optimization Day Hosted by: Department of Management Sciences, |
Organized by:
Samir Elhedhli, Ph.D.,
Associate Professor,
Department of Management Sciences,
elhedhli-at-uwaterloo.ca
8:30 |
_ |
9:00 |
Breakfast---Opening Remarks
by Mike Worswick |
||||
9:00 |
_ |
9:25 |
|
Presentation 1 |
Prof. David Fuller |
MSCI |
A
Large Scale Multi-Player Optimization Model to Assess Market Power in a Time
of Use Electricity Market |
9:25 |
_ |
9:50 |
|
Presentation 2 |
Prof. Lee Fu |
CIVE |
Optimization
under Uncertainty with Applications in Intelligent Transportation Systems |
9:50 |
_ |
10:15 |
|
Presentation 3 |
Prof. S. Elhedhli |
MSCI |
Large-scale optimization:
Applications and solution approaches. |
10:15 |
_ |
10:40 |
|
Presentation 4 |
Prof. James Craig |
CIVE |
Optimal
Management of Groundwater-Surface Water Resources |
10:40 |
_ |
11:55 |
Coffee break |
||||
11:00 |
_ |
11:25 |
|
Presentation 5 |
Prof. Beth. Jewkes |
MSCI |
Optimization
Using Matrix Geometric and Cutting Plane Methods |
|
|
|
|
||||
11:25 |
_ |
11:50 |
|
Presentation 6 |
Masoud Mahootchi |
SYDE |
Reservoir
Management with Reinforcement Learning |
|
|
|
Prof. P. Ponnambalam &
Prof. Hamid Tizhoosh |
||||
11:50 |
_ |
12:15 |
|
Presentation 7 |
Rafael Avalos |
ECE/MSCI |
Solving
Voltage-Stability-Constrained Optimal Power Flows Using a Singular Value Decomposition
Technique |
|
|
|
Prof. Miguel Anjos & Prof. Claudio Canizares |
||||
12:15 |
_ |
12:40 |
|
Presentation 8 |
Yasser Mustafa Prof. Ehab El-Saadany |
ECE |
Optimum size and location
of wind based DG in distribution systems |
12:40 |
_ |
1:35 |
Lunch break |
||||
|
|
|
|
Presentation 9 |
Hatem ElBehairy |
CIVE |
Improving
the Performance of Evolutionary Algorithms for Large-Scale Optimization of
Bridge Deck Maintenance |
1:35 |
_ |
2:00 |
Prof. Tarek Hegazy |
||||
|
|
|
Prof. Khaled Soudki |
||||
2:00 |
_ |
2:25 |
|
Presentation 10 |
Christina Moulton |
SYDE |
Post-Pareto Analysis of
Evolutionary Multiobjective Programming Results for Landscape Ecology
Planning |
|
|
|
Prof. Paul Calamai |
||||
2:25 |
_ |
2:50 |
|
Presentation 11 |
Ismail El-Samahy |
ECE |
Provision of Reactive Power
Ancillary Services in Deregulated Electricity Markets |
|
|
|
Prof. Kankar Bhattacharya & Prof. Claudio
Canizares |
||||
2:50 |
_ |
3:15 |
|
Presentation 12 |
Sayyid Hasan Riyaz |
ECE |
Span Indexed Greedy Algorithm
for Task Planning in Multi Agent Systems |
|
|
|
Prof. Otman Basir |
||||
3:15 |
_ |
3:30 |
Coffee break |
||||
|
|
|
|
Presentation 13 |
Abbas Ahmadi |
ECE |
A new approach for
clustering data using |
3:30 |
_ |
3:55 |
Prof. Fakhri Karray &
Prof. M. Kamel |
multiple cooperative swarms |
|||
|
|
|
|
Presentation 14 |
M. Ali Ulku |
MSCI |
Optimal Pricing and consolidation cycle for a
shipper |
3:55 |
_ |
4:20 |
Prof. Jim Bookbinder |
||||
|
|
|
|
Presentation 15 |
Youssef Saif |
CHE |
Optimal Design of Hybrid
Membrane Networks for Wastewater Treatment and Minimization |
4:20 |
_ |
4:45 |
Prof. Ali Elkamel &
Prof. M. Pritzker |
||||
|
|
|
|
Presentation 16 |
Alice Malisia |
SYDE |
Applying
Opposition-Based Ideas to Ant Colony Algorithms |
4:45 |
_ |
5:10 |
Prof. Hamid Tizhoosh |
Presentation 1: |
A Large Scale Multi-Player
Optimization Model to Assess Market Power in a Time of Use Electricity Market D.
Fuller Department o f Management Sciences The talk will outline a large scale model that represents the possible behaviour of several large generating companies in a "bilateral" market, on a transmission network, with time of use pricing. The mathematical form of the model is a blend of Nash-Cournot and multi-commodity ideas, leading to a formulation as a variational inequality problem (VIP). VIPs are like constrained optimization models, in that they have feasible regions defined by constraints, but unlike optimization models, there is not a single objective function. In this application, there is a profit objective function for each generating company. Because of the connection to constrained optimization, it has been possible to extend the classical Dantzig-Wolfe decomposition method. |
|
Presentation 2: |
Optimization under Uncertainty with Applications in Intelligent Transportation Systems Lee Fu Department o f Civil and Environmental Engineering In this presentation, I will provide an overview of
several optimization problems that I have been involved over the past 10
years. Most of these problems are related to the specific
transportation engineering field called Intelligent Transportation Systems
(ITS), which is aimed at improving the management and operations of the
existing transportation facilities by applying advanced sensors,
telecommunication, and computer technologies. Example problems include vehicle routing and scheduling in dynamic and stochastic networks, real-time transit operation control, locating of traffic changeable measure signs, and scheduling of winter road maintenance vehicles. |
|
Presentation 3: |
Large-scale optimization:
Applications and solution approaches
Samir. Elhedhli Department o f Management Sciences Real-life optimization
problems are typically large in size. Apart from linear programming, mixed
and nonlinear programming problems could not be solved directly using
commercial solvers. Fortunately, large-scale problems often exhibit special
structure that can be exploited using decomposition approaches. In this talk,
we briefly discuss the main decomposition principle and then discuss a number
of applications in logistics
network |
|
Presentation 4: |
Optimal Management of Groundwater-Surface Water Resources
Department o f Civil and Environmental Engineering
|
|
Presentation 5: |
Optimization
Using Matrix Geometric and Cutting Plane Methods Beth. Jewkes Department o f Management Sciences The
Matrix geometric method often proves very efficient in evaluating complex
queuing systems. However, because it is a computational method,
the performance measures obtained using matrix geometric method often
restricts its utility in optimization models. We present a general framework
for combining the matrix geometric approach with mathematical programming
and an iterative cutting plane algorithm. Constraints to be satisfied in
the problem are identified iteratively from results of the matrix geometric
evaluation of the proposed system alternative, which are then added to the
mathematical model for re-optimization until optimal solution is obtained. We
illustrate the application of the proposed method to a problem of product
pricing, capacity selection and delivery time setting in price and time
sensitive markets, subject to service level constraints. The proposed
approach is shown to work very efficiently, and solves the problem in a few
iterations. |
|
Presentation 6: |
Reservoir Management with Reinforcement Learning Masoud
Mahootchi , P. Ponnambalam
& H. Tizhoosh Department
of Systems Design Engineering Finding
optimal or near-optimal solutions using large-scale optimization models for
real-world problems is a problematic and challenging issue. Our focus here is
to achieve optimal operations of surface water resources including multiple
reservoirs in a unified system for short-term or long-term periods. This is a
crucial and important problem because a system of reservoirs could
potentially be sources for varieties of purposes such as drinking water, power
generation, irrigation, flood control, recreation, navigation, and so on and
water resources becoming more scarce. However, considering all various kinds
of complexities, including stochasticity of the input parameters and
nonlinearity/nonconvexity of objective functions and constraints makes the
traditional optimizations methods inefficient in terms of computing time what
is generally known in the literature on reservoir management as “the curse of
dimensionality” To
tackle this main imperfection, we have applied a simulation-based
optimization technique called Q-Learning. In order to speed up the
convergence and increase the robustness of the Q-Learning method, we have
also developed and implemented a new version of the opposition-based
learning. We have used some stochastic optimization models to evaluate the
efficiency of the method. |
|
Presentation 7: |
Solving Voltage-Stability-Constrained Optimal
Power Flows Using a Singular Value
Decomposition Technique Rafael Avalos, Miguel
Anjos & Claudio Canizares Department o f Management Sciences & Department
o f Electrical and Computer Engineering This work proposes a method to enforce a voltage stability constraint in an Optimal Power Flow using the minimum singular value and minimum singular vectors of the power flow Jacobian. By ensuring this singular value is bounded away from zero, the voltage collapse point of the power system is avoided. A singular value decomposition of the power flow Jacobian at the current solution is carried out in an iterative process until the constraint is satisfied. The results show that this method can be a good alternative to other methods previously proposed for this problem. |
|
Presentation 8: |
Optimum size and location of wind based DG in distribution
system Yasser Mustafa, M.M. Salama & . Ehab El-Saadany Department o f Electrical and Computer Engineering DG often offers a valuable
alternative to traditional sources of electrical power for industrial,
commercial, and residential loads, particularly, for rural areas where long
transmission lines are used to transfer the electrical power from main source
to end users. In this work an optimal power flow (OPF) algorithm is used to
determine the optimum allocation of wind based DGs in order to minimizing the
total power losses in the system. The OPF is formulated as Mixed Integer Non
Linear Programming (MINLP); with an objective function to minimize the system
power losses. The constraints include voltage limits at different buses
(slack, generation, and load buses) of the system, feeder capacity, and
maximum penetration limit of DGs. Moreover, the capacity factor of wind based
DGs is calculated using Rayliegh pdf by the aid of the data of mean wind
speed at the DGs locations and wind turbines data. This proposed OPF is
applied to a typical unbalanced Hydro One system with different scenarios
including restrictions on the sizes and locations of DGs. The results show
that the implementation of DGs has a great effect on the power losses
reduction and scenario5b is the optimal case over other scenarios. |
|
Presentation 9: |
Improving the Performance of Evolutionary Algorithms for Large-Scale
Optimization of Bridge Deck Maintenance Hatem ElBehairy,
Tarek Hegazy & Khaled Soudki Department o f Civil and
Environmental Engineering Bridges are vital links in
infrastructure road networks and require frequent maintenance and repair in
order to keep them functional throughout their service lives. Bridge decks
are considered the most vulnerable component of bridges as they are exposed
to harsh environment and increasing traffic volume. Prioritization of bridge decks for
maintenance and repair purposes involves decisions that bring the highest
return on the repair budget. These decisions, however, represent complex
optimization problems that traditional optimization techniques are often
unable to solve. This research introduces an integrated model for bridge deck
repairs with detailed life cycle costs of network-level and project-level decisions.
Two evolutionary-based optimization techniques, namely genetic algorithms and
shuffled-frog-leaping are then applied on the model in order to optimize
maintenance and repair decisions. Results of both techniques are compared on
case study problems with different numbers of bridges. To improve their
performance on large scale problems, various experiments were conducted with
different objective functions, model formulations, parameter settings, and
pre/post processing functions. The best optimization strategy for this
typical infrastructure problem was determined to be a year-by-year
optimization strategy, coupled with the use of a pre-processing function to
allocate repair funds first to critical bridges. |
|
Presentation 10: |
Post-Pareto Analysis of Evolutionary Multiobjective Programming
Results for Landscape Ecology Planning Christina Moulton, Paul Calamai, & Steven Roberts Department of Systems Design Engineering Multiobjective
genetic algorithms such as the Non-dominated Sorting Genetic Algorithm II
(Deb, Pratap, Agarak, & Meyarivan, 2002) can generate an even
distribution of points over the Pareto front (Mattson &
Messac, 2005). Many real world applications involve many objective
function and the Pareto front may contain a very large number of
points. Choosing a solution from such a large set is potentially
intractable for a decision maker (Purhouse & Fleming, 2003). Cluster
analysis can be used to group the Pareto front into subsets of similar
solutions. This approach is applied to a land use change problem in an
urban fringe area in Southern Ontario, Canada. Using NSGA we generate
potential land use plans which are Pareto optimal on 8 landscape ecology
metrics. These plans are clustered based on their performance on the
objective functions. |
|
Presentation 11: |
Provision of Reactive Power Ancillary Services in Deregulated
Electricity Markets Ismail El-Samahy, Kankar
Bhattacharya & Claudio Canizares Department o f Electrical and Computer Engineering Power
system operators in deregulated electricity markets are required to make
provisions for reactive power services in order to maintain an adequate level
of grid security and reliability. Consequently, there is a need to devise
appropriate payment mechanisms for such services, considering their effect on
system operating conditions. These payments should truly reflect the cost
associated with reactive power production, which depends on the region of
operation of the synchronous generator as a reactive power service provider.
Thus, the optimization problem for service procurement should be modeled in a
way that assigns, for each generator, only one of the operating regions. Formulating
the procurement problem as a mixed integer non-linear programming (MINLP)
problem has been previously proposed in literature. However, it is
computationally complex to augment this model with security constraints such
as thermal limits on transmission lines. Thus, a three-step iterative scheme
based on the classification of reactive power from a generator is proposed in
this work to eliminate the integer constraints, and hence reformulate the
problem as a non-linear programming (NLP) instead of a MINLP problem. The
proposed algorithm reduces the computational burden and allows for the
inclusion of essential security constraints. |
|
Presentation 12: |
Span Indexed Greedy
Algorithm for Task Planning in Multi Agent Systems Sayyid Hasan Riyaz and
Otman Basir Department o f Electrical and Computer Engineering Multi Agent Systems (MAS) are often employed in applications having varying and uncertain dynamics resulting from a number of sources such as operational environment, agent attributes and task dynamics. Traditional static task planning optimization techniques often prove infeasible for such systems and alternate dynamic techniques are considered that are responsive to the variations and uncertainties of the system. Market mechanisms based techniques using greedy algorithm have been a popular choice for MAS task planning due to their
simplicity and effectiveness. Solutions based on greedy algorithm, although
feasible, are not guaranteed to be optimal thus providing an opportunity
of improvement in the solution quality. We propose a novel Span Indexed
Greedy Algorithm (SIGA) for dynamic task scheduling and allocation in MAS and
demonstrate its efficiency and efficacy through comparative studies and
analysis with the greedy algorithm and the optimal solutions. |
|
Presentation 13: |
A new approach for clustering data using multiple cooperative swarms Abbas Ahmadi , Fakhri Karray, Mohamed
Kamel Department o f Electrical and Computer Engineering Particle swarm
optimization (PSO) is a search technique which is mainly introduced to deal
with optimization problems by defining an objective function, called fitness
function. It starts from an initial population and explores the feasible
solution region, using several particles through a number of iterations to
reach a near optimal solution. Due to its abilities, PSO has been applied in
other applications such as classification and clustering. Current PSO-based
techniques have been developed using a single swarm. We
propose a new clustering technique using multiple swarms. This technique
imitates the behavior of biological swarms which search for food situated in
several places. The proposed technique considers multiple cooperating swarms
to find centers of clusters. This
task is done during two phases: initialization and exploration. In the
initialization phase, it assigns a portion of the feasible solution region to
each swarm. In the exploration phase, the search to find the near optimal
solution proceeds using cooperating swarms. As compared to the
single-swarm-based clustering technique, multiple cooperating swarms provide
more effective solutions for the clustering task as the dimensionality of
data and number of clusters increase. The proposed technique outperforms k-means
clustering and single-swarm-based clustering techniques. |
|
Presentation 14: |
Optimal
Pricing and consolidation cycle for a shipper M. Ali Ulku & Jim Bookbinder Department o f Management Sciences We consider a shipper
whose customers are sensitive to both price and delivery-time guarantee. For
various pricing schemes and via classical optimization methods, we study the
problem of maximizing the shipper's profit per unit time through the optimal
choice of consolidation cycle length, when a contract carrier is employed. We
show that the problem is very parameter-sensitive, and the best choice of
pricing scheme differs with management's particular objective. Moreover, we
investigate the market conditions in which shipment consolidation might not
be preferred at all. Contrary to intuition, it turns out that charging
according to an order's time of arrival is not necessarily the best pricing
scheme. We provide managerial insights and include a numerical example with various
sensitivity analyses. Keywords: pricing, shipment consolidation,
logistics, price-and delivery-time sensitiveness |
|
Presentation 15: |
Optimal Design of Hybrid
Membrane Networks for Wastewater Treatment and Minimization Y. Saif, A. Elkamel and M. Pritzker Department of Chemical
Engineering The current research addresses mainly
deterministic optimization problems of water desalination and industrial
wastewater treatment networks by hybrid membranes. The mathematical
optimization problem involves the solution of a nonconvex mixed integer
nonlinear system of equations (MINLP). The mathematical models are formulated
with the view of creating general superstructures which give a rich number of
alternatives for an optimum treatment network. Several case studies will be
tackled to highlight the reliability and efficiency of membranes related to
water and wastewater treatments. A global optimization method is proposed to
solve the membrane network based on the reformulation-linearization technique
(RLT). Preliminary results are presented to illustrate the feasibility of the
solution methodology. |
|
Presentation 16: |
Applying Opposition-Based Ideas to Ant
Algorithms Alice R. Malisia and
Hamid R. Tizhoosh Department of Systems Design Engineering The concept of opposition-based learning (OBL) was
recently proposed by Dr. Tizhoosh to extend different machine learning
algorithms. The main idea of OBL is to consider opposite estimates, actions or
states as an attempt to increase the coverage of the solution space and to
reduce exploration time. OBL has already been applied to reinforcement
learning, neural networks and genetic algorithms. Our goal is to apply OBL to ant algorithms. This presentation
will provide an overview of the investigations that have been conducted. We
will present five different extensions applied to a particular version of ant
algorithms: ant colony system (ACS). The modifications focus on the solution
construction phase of the ant colony system. Three of the proposed methods
work by pairing the ants and synchronizing their path selection. The two
other approaches modify the decisions of the ants by using an
opposite-pheromone content. Results on the application of these algorithms to
Travelling Salesman Problem instances demonstrate that the concept of
opposition is not easily applied to the ant algorithm. Only one of the
pheromone-based methods showed performance improvements that were
statistically significant. The quality of the solutions increased and more
optimal solutions were found. The other extensions showed no clear
improvement. Further work must
be conducted to explore the more successful pheromone-based approach. Additionally,
we began experimenting with a different application: search of a goal within
a grid environment. Moreover, it would be important to establish the benefits
and limitations of applying opposition to ant algorithms. |
|